1 | #include "EXTERN.h" |
1 | #include "EXTERN.h" |
2 | #include "perl.h" |
2 | #include "perl.h" |
3 | #include "XSUB.h" |
3 | #include "XSUB.h" |
4 | |
4 | |
5 | #include "assert.h" |
5 | #include <assert.h> |
6 | #include "string.h" |
6 | #include <string.h> |
7 | #include "stdlib.h" |
7 | #include <stdlib.h> |
|
|
8 | #include <stdio.h> |
|
|
9 | #include <float.h> |
8 | |
10 | |
|
|
11 | #if defined(__BORLANDC__) || defined(_MSC_VER) |
|
|
12 | # define snprintf _snprintf // C compilers have this in stdio.h |
|
|
13 | #endif |
|
|
14 | |
|
|
15 | // some old perls do not have this, try to make it work, no |
|
|
16 | // guarentees, though. if it breaks, you get to keep the pieces. |
|
|
17 | #ifndef UTF8_MAXBYTES |
|
|
18 | # define UTF8_MAXBYTES 13 |
|
|
19 | #endif |
|
|
20 | |
9 | #define F_ASCII 0x00000001 |
21 | #define F_ASCII 0x00000001UL |
|
|
22 | #define F_LATIN1 0x00000002UL |
10 | #define F_UTF8 0x00000002 |
23 | #define F_UTF8 0x00000004UL |
11 | #define F_INDENT 0x00000004 |
24 | #define F_INDENT 0x00000008UL |
12 | #define F_CANONICAL 0x00000008 |
25 | #define F_CANONICAL 0x00000010UL |
13 | #define F_SPACE_BEFORE 0x00000010 |
26 | #define F_SPACE_BEFORE 0x00000020UL |
14 | #define F_SPACE_AFTER 0x00000020 |
27 | #define F_SPACE_AFTER 0x00000040UL |
15 | #define F_JSON_RPC 0x00000040 |
|
|
16 | #define F_ALLOW_NONREF 0x00000080 |
28 | #define F_ALLOW_NONREF 0x00000100UL |
|
|
29 | #define F_SHRINK 0x00000200UL |
|
|
30 | #define F_ALLOW_BLESSED 0x00000400UL |
|
|
31 | #define F_CONV_BLESSED 0x00000800UL |
|
|
32 | #define F_RELAXED 0x00001000UL |
|
|
33 | |
|
|
34 | #define F_MAXDEPTH 0xf8000000UL |
|
|
35 | #define S_MAXDEPTH 27 |
|
|
36 | #define F_MAXSIZE 0x01f00000UL |
|
|
37 | #define S_MAXSIZE 20 |
|
|
38 | #define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing |
|
|
39 | |
|
|
40 | #define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH)) |
|
|
41 | #define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE )) |
17 | |
42 | |
18 | #define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER |
43 | #define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER |
19 | #define F_DEFAULT 0 |
44 | #define F_DEFAULT (9UL << S_MAXDEPTH) |
20 | |
45 | |
21 | #define INIT_SIZE 32 // initial scalar size to be allocated |
46 | #define INIT_SIZE 32 // initial scalar size to be allocated |
|
|
47 | #define INDENT_STEP 3 // spaces per indentation level |
|
|
48 | |
|
|
49 | #define SHORT_STRING_LEN 16384 // special-case strings of up to this size |
22 | |
50 | |
23 | #define SB do { |
51 | #define SB do { |
24 | #define SE } while (0) |
52 | #define SE } while (0) |
25 | |
53 | |
26 | static HV *json_stash; |
54 | #if __GNUC__ >= 3 |
|
|
55 | # define expect(expr,value) __builtin_expect ((expr),(value)) |
|
|
56 | # define inline inline |
|
|
57 | #else |
|
|
58 | # define expect(expr,value) (expr) |
|
|
59 | # define inline static |
|
|
60 | #endif |
|
|
61 | |
|
|
62 | #define expect_false(expr) expect ((expr) != 0, 0) |
|
|
63 | #define expect_true(expr) expect ((expr) != 0, 1) |
|
|
64 | |
|
|
65 | #ifdef USE_ITHREADS |
|
|
66 | # define JSON_SLOW 1 |
|
|
67 | # define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1)) |
|
|
68 | #else |
|
|
69 | # define JSON_SLOW 0 |
|
|
70 | # define JSON_STASH json_stash |
|
|
71 | #endif |
|
|
72 | |
|
|
73 | static HV *json_stash, *json_boolean_stash; // JSON::XS:: |
|
|
74 | static SV *json_true, *json_false; |
|
|
75 | |
|
|
76 | typedef struct { |
|
|
77 | U32 flags; |
|
|
78 | SV *cb_object; |
|
|
79 | HV *cb_sk_object; |
|
|
80 | } JSON; |
|
|
81 | |
|
|
82 | ///////////////////////////////////////////////////////////////////////////// |
|
|
83 | // utility functions |
|
|
84 | |
|
|
85 | inline void |
|
|
86 | shrink (SV *sv) |
|
|
87 | { |
|
|
88 | sv_utf8_downgrade (sv, 1); |
|
|
89 | if (SvLEN (sv) > SvCUR (sv) + 1) |
|
|
90 | { |
|
|
91 | #ifdef SvPV_shrink_to_cur |
|
|
92 | SvPV_shrink_to_cur (sv); |
|
|
93 | #elif defined (SvPV_renew) |
|
|
94 | SvPV_renew (sv, SvCUR (sv) + 1); |
|
|
95 | #endif |
|
|
96 | } |
|
|
97 | } |
|
|
98 | |
|
|
99 | // decode an utf-8 character and return it, or (UV)-1 in |
|
|
100 | // case of an error. |
|
|
101 | // we special-case "safe" characters from U+80 .. U+7FF, |
|
|
102 | // but use the very good perl function to parse anything else. |
|
|
103 | // note that we never call this function for a ascii codepoints |
|
|
104 | inline UV |
|
|
105 | decode_utf8 (unsigned char *s, STRLEN len, STRLEN *clen) |
|
|
106 | { |
|
|
107 | if (expect_false (s[0] > 0xdf || s[0] < 0xc2)) |
|
|
108 | return utf8n_to_uvuni (s, len, clen, UTF8_CHECK_ONLY); |
|
|
109 | else if (len > 1 && s[1] >= 0x80 && s[1] <= 0xbf) |
|
|
110 | { |
|
|
111 | *clen = 2; |
|
|
112 | return ((s[0] & 0x1f) << 6) | (s[1] & 0x3f); |
|
|
113 | } |
|
|
114 | else |
|
|
115 | { |
|
|
116 | *clen = (STRLEN)-1; |
|
|
117 | return (UV)-1; |
|
|
118 | } |
|
|
119 | } |
|
|
120 | |
|
|
121 | ///////////////////////////////////////////////////////////////////////////// |
|
|
122 | // encoder |
27 | |
123 | |
28 | // structure used for encoding JSON |
124 | // structure used for encoding JSON |
29 | typedef struct |
125 | typedef struct |
30 | { |
126 | { |
31 | char *cur; |
127 | char *cur; // SvPVX (sv) + current output position |
32 | STRLEN len; // SvLEN (sv) |
|
|
33 | char *end; // SvEND (sv) |
128 | char *end; // SvEND (sv) |
34 | SV *sv; |
129 | SV *sv; // result scalar |
35 | UV flags; |
130 | JSON json; |
36 | int max_recurse; |
131 | U32 indent; // indentation level |
37 | int indent; |
132 | U32 maxdepth; // max. indentation/recursion level |
38 | } enc_t; |
133 | } enc_t; |
39 | |
134 | |
40 | // structure used for decoding JSON |
135 | inline void |
41 | typedef struct |
|
|
42 | { |
|
|
43 | char *cur; |
|
|
44 | char *end; |
|
|
45 | const char *err; |
|
|
46 | UV flags; |
|
|
47 | } dec_t; |
|
|
48 | |
|
|
49 | static UV * |
|
|
50 | SvJSON (SV *sv) |
|
|
51 | { |
|
|
52 | if (!(SvROK (sv) && SvOBJECT (SvRV (sv)) && SvSTASH (SvRV (sv)) == json_stash)) |
|
|
53 | croak ("object is not of type JSON::XS"); |
|
|
54 | |
|
|
55 | return &SvUVX (SvRV (sv)); |
|
|
56 | } |
|
|
57 | |
|
|
58 | ///////////////////////////////////////////////////////////////////////////// |
|
|
59 | |
|
|
60 | static void |
|
|
61 | need (enc_t *enc, STRLEN len) |
136 | need (enc_t *enc, STRLEN len) |
62 | { |
137 | { |
63 | if (enc->cur + len >= enc->end) |
138 | if (expect_false (enc->cur + len >= enc->end)) |
64 | { |
139 | { |
65 | STRLEN cur = enc->cur - SvPVX (enc->sv); |
140 | STRLEN cur = enc->cur - SvPVX (enc->sv); |
66 | SvGROW (enc->sv, cur + len + 1); |
141 | SvGROW (enc->sv, cur + len + 1); |
67 | enc->cur = SvPVX (enc->sv) + cur; |
142 | enc->cur = SvPVX (enc->sv) + cur; |
68 | enc->end = SvPVX (enc->sv) + SvLEN (enc->sv); |
143 | enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; |
69 | } |
144 | } |
70 | } |
145 | } |
71 | |
146 | |
72 | static void |
147 | inline void |
73 | encode_ch (enc_t *enc, char ch) |
148 | encode_ch (enc_t *enc, char ch) |
74 | { |
149 | { |
75 | need (enc, 1); |
150 | need (enc, 1); |
76 | *enc->cur++ = ch; |
151 | *enc->cur++ = ch; |
77 | } |
152 | } |
… | |
… | |
85 | |
160 | |
86 | while (str < end) |
161 | while (str < end) |
87 | { |
162 | { |
88 | unsigned char ch = *(unsigned char *)str; |
163 | unsigned char ch = *(unsigned char *)str; |
89 | |
164 | |
90 | if (ch == '"') |
165 | if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case |
91 | { |
166 | { |
|
|
167 | if (expect_false (ch == '"')) // but with slow exceptions |
|
|
168 | { |
92 | need (enc, len += 1); |
169 | need (enc, len += 1); |
93 | *enc->cur++ = '\\'; |
170 | *enc->cur++ = '\\'; |
94 | *enc->cur++ = '"'; |
171 | *enc->cur++ = '"'; |
95 | ++str; |
|
|
96 | } |
172 | } |
97 | else if (ch == '\\') |
173 | else if (expect_false (ch == '\\')) |
98 | { |
174 | { |
99 | need (enc, len += 1); |
175 | need (enc, len += 1); |
100 | *enc->cur++ = '\\'; |
176 | *enc->cur++ = '\\'; |
101 | *enc->cur++ = '\\'; |
177 | *enc->cur++ = '\\'; |
102 | ++str; |
|
|
103 | } |
178 | } |
104 | else if (ch >= 0x20 && ch < 0x80) // most common case |
179 | else |
105 | { |
|
|
106 | *enc->cur++ = ch; |
180 | *enc->cur++ = ch; |
107 | ++str; |
181 | |
108 | } |
|
|
109 | else if (ch == '\015') |
|
|
110 | { |
|
|
111 | need (enc, len += 1); |
|
|
112 | *enc->cur++ = '\\'; |
|
|
113 | *enc->cur++ = 'r'; |
|
|
114 | ++str; |
|
|
115 | } |
|
|
116 | else if (ch == '\012') |
|
|
117 | { |
|
|
118 | need (enc, len += 1); |
|
|
119 | *enc->cur++ = '\\'; |
|
|
120 | *enc->cur++ = 'n'; |
|
|
121 | ++str; |
182 | ++str; |
122 | } |
183 | } |
123 | else |
184 | else |
124 | { |
185 | { |
125 | STRLEN clen; |
186 | switch (ch) |
126 | UV uch; |
|
|
127 | |
|
|
128 | if (is_utf8) |
|
|
129 | { |
|
|
130 | uch = utf8n_to_uvuni (str, end - str, &clen, UTF8_CHECK_ONLY); |
|
|
131 | if (clen == (STRLEN)-1) |
|
|
132 | croak ("malformed UTF-8 character in string, cannot convert to JSON"); |
|
|
133 | } |
187 | { |
134 | else |
188 | case '\010': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'b'; ++str; break; |
135 | { |
189 | case '\011': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 't'; ++str; break; |
136 | uch = ch; |
190 | case '\012': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'n'; ++str; break; |
137 | clen = 1; |
191 | case '\014': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'f'; ++str; break; |
138 | } |
192 | case '\015': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'r'; ++str; break; |
139 | |
193 | |
140 | if (uch < 0x80 || enc->flags & F_ASCII) |
194 | default: |
141 | { |
|
|
142 | if (uch > 0xFFFFUL) |
|
|
143 | { |
195 | { |
|
|
196 | STRLEN clen; |
|
|
197 | UV uch; |
|
|
198 | |
|
|
199 | if (is_utf8) |
|
|
200 | { |
|
|
201 | uch = decode_utf8 (str, end - str, &clen); |
|
|
202 | if (clen == (STRLEN)-1) |
|
|
203 | croak ("malformed or illegal unicode character in string [%.11s], cannot convert to JSON", str); |
|
|
204 | } |
|
|
205 | else |
|
|
206 | { |
|
|
207 | uch = ch; |
|
|
208 | clen = 1; |
|
|
209 | } |
|
|
210 | |
|
|
211 | if (uch > 0x10FFFFUL) |
|
|
212 | croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch); |
|
|
213 | |
|
|
214 | if (uch < 0x80 || enc->json.flags & F_ASCII || (enc->json.flags & F_LATIN1 && uch > 0xFF)) |
|
|
215 | { |
|
|
216 | if (uch > 0xFFFFUL) |
|
|
217 | { |
144 | need (enc, len += 11); |
218 | need (enc, len += 11); |
145 | sprintf (enc->cur, "\\u%04x\\u%04x", |
219 | sprintf (enc->cur, "\\u%04x\\u%04x", |
146 | (uch - 0x10000) / 0x400 + 0xD800, |
220 | (int)((uch - 0x10000) / 0x400 + 0xD800), |
147 | (uch - 0x10000) % 0x400 + 0xDC00); |
221 | (int)((uch - 0x10000) % 0x400 + 0xDC00)); |
148 | enc->cur += 12; |
222 | enc->cur += 12; |
|
|
223 | } |
|
|
224 | else |
|
|
225 | { |
|
|
226 | static char hexdigit [16] = "0123456789abcdef"; |
|
|
227 | need (enc, len += 5); |
|
|
228 | *enc->cur++ = '\\'; |
|
|
229 | *enc->cur++ = 'u'; |
|
|
230 | *enc->cur++ = hexdigit [ uch >> 12 ]; |
|
|
231 | *enc->cur++ = hexdigit [(uch >> 8) & 15]; |
|
|
232 | *enc->cur++ = hexdigit [(uch >> 4) & 15]; |
|
|
233 | *enc->cur++ = hexdigit [(uch >> 0) & 15]; |
|
|
234 | } |
|
|
235 | |
|
|
236 | str += clen; |
|
|
237 | } |
|
|
238 | else if (enc->json.flags & F_LATIN1) |
|
|
239 | { |
|
|
240 | *enc->cur++ = uch; |
|
|
241 | str += clen; |
|
|
242 | } |
|
|
243 | else if (is_utf8) |
|
|
244 | { |
|
|
245 | need (enc, len += clen); |
|
|
246 | do |
|
|
247 | { |
|
|
248 | *enc->cur++ = *str++; |
|
|
249 | } |
|
|
250 | while (--clen); |
|
|
251 | } |
|
|
252 | else |
|
|
253 | { |
|
|
254 | need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed |
|
|
255 | enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0); |
|
|
256 | ++str; |
|
|
257 | } |
149 | } |
258 | } |
150 | else |
|
|
151 | { |
|
|
152 | static char hexdigit [16] = "0123456789abcdef"; |
|
|
153 | need (enc, len += 5); |
|
|
154 | *enc->cur++ = '\\'; |
|
|
155 | *enc->cur++ = 'u'; |
|
|
156 | *enc->cur++ = hexdigit [ uch >> 12 ]; |
|
|
157 | *enc->cur++ = hexdigit [(uch >> 8) & 15]; |
|
|
158 | *enc->cur++ = hexdigit [(uch >> 4) & 15]; |
|
|
159 | *enc->cur++ = hexdigit [(uch >> 0) & 15]; |
|
|
160 | } |
|
|
161 | |
|
|
162 | str += clen; |
|
|
163 | } |
|
|
164 | else if (is_utf8) |
|
|
165 | { |
|
|
166 | need (enc, len += clen); |
|
|
167 | do |
|
|
168 | { |
|
|
169 | *enc->cur++ = *str++; |
|
|
170 | } |
|
|
171 | while (--clen); |
|
|
172 | } |
|
|
173 | else |
|
|
174 | { |
|
|
175 | need (enc, 10); // never more than 11 bytes needed |
|
|
176 | enc->cur = uvuni_to_utf8_flags (enc->cur, uch, 0); |
|
|
177 | ++str; |
|
|
178 | } |
259 | } |
179 | } |
260 | } |
180 | |
261 | |
181 | --len; |
262 | --len; |
182 | } |
263 | } |
183 | } |
264 | } |
184 | |
265 | |
185 | #define INDENT SB \ |
266 | inline void |
|
|
267 | encode_indent (enc_t *enc) |
|
|
268 | { |
186 | if (enc->flags & F_INDENT) \ |
269 | if (enc->json.flags & F_INDENT) |
187 | { \ |
270 | { |
188 | int i_; \ |
271 | int spaces = enc->indent * INDENT_STEP; |
189 | need (enc, enc->indent); \ |
272 | |
190 | for (i_ = enc->indent * 3; i_--; )\ |
273 | need (enc, spaces); |
|
|
274 | memset (enc->cur, ' ', spaces); |
|
|
275 | enc->cur += spaces; |
|
|
276 | } |
|
|
277 | } |
|
|
278 | |
|
|
279 | inline void |
|
|
280 | encode_space (enc_t *enc) |
|
|
281 | { |
|
|
282 | need (enc, 1); |
|
|
283 | encode_ch (enc, ' '); |
|
|
284 | } |
|
|
285 | |
|
|
286 | inline void |
|
|
287 | encode_nl (enc_t *enc) |
|
|
288 | { |
|
|
289 | if (enc->json.flags & F_INDENT) |
|
|
290 | { |
|
|
291 | need (enc, 1); |
191 | encode_ch (enc, ' '); \ |
292 | encode_ch (enc, '\n'); |
192 | } \ |
293 | } |
193 | SE |
294 | } |
194 | |
295 | |
195 | #define SPACE SB need (enc, 1); encode_ch (enc, ' '); SE |
296 | inline void |
196 | #define NL SB if (enc->flags & F_INDENT) { need (enc, 1); encode_ch (enc, '\n'); } SE |
297 | encode_comma (enc_t *enc) |
197 | #define COMMA SB \ |
298 | { |
198 | encode_ch (enc, ','); \ |
299 | encode_ch (enc, ','); |
|
|
300 | |
199 | if (enc->flags & F_INDENT) \ |
301 | if (enc->json.flags & F_INDENT) |
200 | NL; \ |
302 | encode_nl (enc); |
201 | else if (enc->flags & F_SPACE_AFTER) \ |
303 | else if (enc->json.flags & F_SPACE_AFTER) |
202 | SPACE; \ |
304 | encode_space (enc); |
203 | SE |
305 | } |
204 | |
306 | |
205 | static void encode_sv (enc_t *enc, SV *sv); |
307 | static void encode_sv (enc_t *enc, SV *sv); |
206 | |
308 | |
207 | static void |
309 | static void |
208 | encode_av (enc_t *enc, AV *av) |
310 | encode_av (enc_t *enc, AV *av) |
209 | { |
311 | { |
210 | int i, len = av_len (av); |
312 | int i, len = av_len (av); |
211 | |
313 | |
212 | encode_ch (enc, '['); NL; |
314 | if (enc->indent >= enc->maxdepth) |
|
|
315 | croak ("data structure too deep (hit recursion limit)"); |
|
|
316 | |
|
|
317 | encode_ch (enc, '['); encode_nl (enc); |
213 | ++enc->indent; |
318 | ++enc->indent; |
214 | |
319 | |
215 | for (i = 0; i <= len; ++i) |
320 | for (i = 0; i <= len; ++i) |
216 | { |
321 | { |
217 | INDENT; |
322 | SV **svp = av_fetch (av, i, 0); |
218 | encode_sv (enc, *av_fetch (av, i, 0)); |
323 | |
|
|
324 | encode_indent (enc); |
|
|
325 | |
|
|
326 | if (svp) |
|
|
327 | encode_sv (enc, *svp); |
|
|
328 | else |
|
|
329 | encode_str (enc, "null", 4, 0); |
219 | |
330 | |
220 | if (i < len) |
331 | if (i < len) |
221 | COMMA; |
332 | encode_comma (enc); |
222 | } |
333 | } |
223 | |
334 | |
224 | NL; |
335 | encode_nl (enc); |
225 | |
336 | |
226 | --enc->indent; |
337 | --enc->indent; |
227 | INDENT; encode_ch (enc, ']'); |
338 | encode_indent (enc); encode_ch (enc, ']'); |
228 | } |
339 | } |
229 | |
340 | |
230 | static void |
341 | static void |
231 | encode_he (enc_t *enc, HE *he) |
342 | encode_hk (enc_t *enc, HE *he) |
232 | { |
343 | { |
233 | encode_ch (enc, '"'); |
344 | encode_ch (enc, '"'); |
234 | |
345 | |
235 | if (HeKLEN (he) == HEf_SVKEY) |
346 | if (HeKLEN (he) == HEf_SVKEY) |
236 | { |
347 | { |
… | |
… | |
246 | else |
357 | else |
247 | encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he)); |
358 | encode_str (enc, HeKEY (he), HeKLEN (he), HeKUTF8 (he)); |
248 | |
359 | |
249 | encode_ch (enc, '"'); |
360 | encode_ch (enc, '"'); |
250 | |
361 | |
251 | if (enc->flags & F_SPACE_BEFORE) SPACE; |
362 | if (enc->json.flags & F_SPACE_BEFORE) encode_space (enc); |
252 | encode_ch (enc, ':'); |
363 | encode_ch (enc, ':'); |
253 | if (enc->flags & F_SPACE_AFTER ) SPACE; |
364 | if (enc->json.flags & F_SPACE_AFTER ) encode_space (enc); |
254 | encode_sv (enc, HeVAL (he)); |
|
|
255 | } |
365 | } |
256 | |
366 | |
257 | // compare hash entries, used when all keys are bytestrings |
367 | // compare hash entries, used when all keys are bytestrings |
258 | static int |
368 | static int |
259 | he_cmp_fast (const void *a_, const void *b_) |
369 | he_cmp_fast (const void *a_, const void *b_) |
… | |
… | |
264 | HE *b = *(HE **)b_; |
374 | HE *b = *(HE **)b_; |
265 | |
375 | |
266 | STRLEN la = HeKLEN (a); |
376 | STRLEN la = HeKLEN (a); |
267 | STRLEN lb = HeKLEN (b); |
377 | STRLEN lb = HeKLEN (b); |
268 | |
378 | |
269 | if (!(cmp == memcmp (HeKEY (a), HeKEY (b), la < lb ? la : lb))) |
379 | if (!(cmp = memcmp (HeKEY (b), HeKEY (a), lb < la ? lb : la))) |
270 | cmp = la < lb ? -1 : la == lb ? 0 : 1; |
380 | cmp = lb - la; |
271 | |
381 | |
272 | return cmp; |
382 | return cmp; |
273 | } |
383 | } |
274 | |
384 | |
275 | // compare hash entries, used when some keys are sv's or utf-x |
385 | // compare hash entries, used when some keys are sv's or utf-x |
276 | static int |
386 | static int |
277 | he_cmp_slow (const void *a, const void *b) |
387 | he_cmp_slow (const void *a, const void *b) |
278 | { |
388 | { |
279 | return sv_cmp (HeSVKEY_force (*(HE **)a), HeSVKEY_force (*(HE **)b)); |
389 | return sv_cmp (HeSVKEY_force (*(HE **)b), HeSVKEY_force (*(HE **)a)); |
280 | } |
390 | } |
281 | |
391 | |
282 | static void |
392 | static void |
283 | encode_hv (enc_t *enc, HV *hv) |
393 | encode_hv (enc_t *enc, HV *hv) |
284 | { |
394 | { |
|
|
395 | HE *he; |
285 | int count, i; |
396 | int count; |
286 | |
397 | |
|
|
398 | if (enc->indent >= enc->maxdepth) |
|
|
399 | croak ("data structure too deep (hit recursion limit)"); |
|
|
400 | |
287 | encode_ch (enc, '{'); NL; ++enc->indent; |
401 | encode_ch (enc, '{'); encode_nl (enc); ++enc->indent; |
288 | |
402 | |
289 | if ((count = hv_iterinit (hv))) |
|
|
290 | { |
|
|
291 | // for canonical output we have to sort by keys first |
403 | // for canonical output we have to sort by keys first |
292 | // actually, this is mostly due to the stupid so-called |
404 | // actually, this is mostly due to the stupid so-called |
293 | // security workaround added somewhere in 5.8.x. |
405 | // security workaround added somewhere in 5.8.x. |
294 | // that randomises hash orderings |
406 | // that randomises hash orderings |
295 | if (enc->flags & F_CANONICAL) |
407 | if (enc->json.flags & F_CANONICAL) |
|
|
408 | { |
|
|
409 | int count = hv_iterinit (hv); |
|
|
410 | |
|
|
411 | if (SvMAGICAL (hv)) |
296 | { |
412 | { |
297 | HE *he, *hes [count]; |
413 | // need to count by iterating. could improve by dynamically building the vector below |
|
|
414 | // but I don't care for the speed of this special case. |
|
|
415 | // note also that we will run into undefined behaviour when the two iterations |
|
|
416 | // do not result in the same count, something I might care for in some later release. |
|
|
417 | |
|
|
418 | count = 0; |
|
|
419 | while (hv_iternext (hv)) |
|
|
420 | ++count; |
|
|
421 | |
|
|
422 | hv_iterinit (hv); |
|
|
423 | } |
|
|
424 | |
|
|
425 | if (count) |
|
|
426 | { |
298 | int fast = 1; |
427 | int i, fast = 1; |
|
|
428 | #if defined(__BORLANDC__) || defined(_MSC_VER) |
|
|
429 | HE **hes = _alloca (count * sizeof (HE)); |
|
|
430 | #else |
|
|
431 | HE *hes [count]; // if your compiler dies here, you need to enable C99 mode |
|
|
432 | #endif |
299 | |
433 | |
300 | i = 0; |
434 | i = 0; |
301 | while ((he = hv_iternext (hv))) |
435 | while ((he = hv_iternext (hv))) |
302 | { |
436 | { |
303 | hes [i++] = he; |
437 | hes [i++] = he; |
… | |
… | |
309 | |
443 | |
310 | if (fast) |
444 | if (fast) |
311 | qsort (hes, count, sizeof (HE *), he_cmp_fast); |
445 | qsort (hes, count, sizeof (HE *), he_cmp_fast); |
312 | else |
446 | else |
313 | { |
447 | { |
314 | // hack to disable "use bytes" |
448 | // hack to forcefully disable "use bytes" |
315 | COP *oldcop = PL_curcop, cop; |
449 | COP cop = *PL_curcop; |
316 | cop.op_private = 0; |
450 | cop.op_private = 0; |
|
|
451 | |
|
|
452 | ENTER; |
|
|
453 | SAVETMPS; |
|
|
454 | |
|
|
455 | SAVEVPTR (PL_curcop); |
317 | PL_curcop = &cop; |
456 | PL_curcop = &cop; |
318 | |
457 | |
319 | SAVETMPS; |
|
|
320 | qsort (hes, count, sizeof (HE *), he_cmp_slow); |
458 | qsort (hes, count, sizeof (HE *), he_cmp_slow); |
|
|
459 | |
321 | FREETMPS; |
460 | FREETMPS; |
322 | |
461 | LEAVE; |
323 | PL_curcop = oldcop; |
|
|
324 | } |
|
|
325 | |
|
|
326 | for (i = 0; i < count; ++i) |
|
|
327 | { |
462 | } |
328 | INDENT; |
|
|
329 | encode_he (enc, hes [i]); |
|
|
330 | |
463 | |
331 | if (i < count - 1) |
464 | while (count--) |
332 | COMMA; |
|
|
333 | } |
465 | { |
|
|
466 | encode_indent (enc); |
|
|
467 | he = hes [count]; |
|
|
468 | encode_hk (enc, he); |
|
|
469 | encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he)); |
334 | |
470 | |
|
|
471 | if (count) |
|
|
472 | encode_comma (enc); |
335 | NL; |
473 | } |
|
|
474 | } |
|
|
475 | } |
|
|
476 | else |
|
|
477 | { |
|
|
478 | if (hv_iterinit (hv) || SvMAGICAL (hv)) |
|
|
479 | if ((he = hv_iternext (hv))) |
|
|
480 | for (;;) |
|
|
481 | { |
|
|
482 | encode_indent (enc); |
|
|
483 | encode_hk (enc, he); |
|
|
484 | encode_sv (enc, expect_false (SvMAGICAL (hv)) ? hv_iterval (hv, he) : HeVAL (he)); |
|
|
485 | |
|
|
486 | if (!(he = hv_iternext (hv))) |
|
|
487 | break; |
|
|
488 | |
|
|
489 | encode_comma (enc); |
|
|
490 | } |
|
|
491 | } |
|
|
492 | |
|
|
493 | encode_nl (enc); |
|
|
494 | |
|
|
495 | --enc->indent; encode_indent (enc); encode_ch (enc, '}'); |
|
|
496 | } |
|
|
497 | |
|
|
498 | // encode objects, arrays and special \0=false and \1=true values. |
|
|
499 | static void |
|
|
500 | encode_rv (enc_t *enc, SV *sv) |
|
|
501 | { |
|
|
502 | svtype svt; |
|
|
503 | |
|
|
504 | SvGETMAGIC (sv); |
|
|
505 | svt = SvTYPE (sv); |
|
|
506 | |
|
|
507 | if (expect_false (SvOBJECT (sv))) |
|
|
508 | { |
|
|
509 | HV *stash = !JSON_SLOW || json_boolean_stash |
|
|
510 | ? json_boolean_stash |
|
|
511 | : gv_stashpv ("JSON::XS::Boolean", 1); |
|
|
512 | |
|
|
513 | if (SvSTASH (sv) == stash) |
|
|
514 | { |
|
|
515 | if (SvIV (sv)) |
|
|
516 | encode_str (enc, "true", 4, 0); |
|
|
517 | else |
|
|
518 | encode_str (enc, "false", 5, 0); |
336 | } |
519 | } |
337 | else |
520 | else |
338 | { |
521 | { |
339 | SV *sv; |
522 | #if 0 |
340 | HE *he = hv_iternext (hv); |
523 | if (0 && sv_derived_from (rv, "JSON::Literal")) |
341 | |
|
|
342 | for (;;) |
|
|
343 | { |
|
|
344 | INDENT; |
|
|
345 | encode_he (enc, he); |
|
|
346 | |
|
|
347 | if (!(he = hv_iternext (hv))) |
|
|
348 | break; |
|
|
349 | |
|
|
350 | COMMA; |
|
|
351 | } |
524 | { |
352 | |
525 | // not yet |
353 | NL; |
526 | } |
|
|
527 | #endif |
|
|
528 | if (enc->json.flags & F_CONV_BLESSED) |
|
|
529 | { |
|
|
530 | // we re-bless the reference to get overload and other niceties right |
|
|
531 | GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0); |
|
|
532 | |
|
|
533 | if (to_json) |
|
|
534 | { |
|
|
535 | dSP; |
|
|
536 | |
|
|
537 | ENTER; SAVETMPS; PUSHMARK (SP); |
|
|
538 | XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv))); |
|
|
539 | |
|
|
540 | // calling with G_SCALAR ensures that we always get a 1 return value |
|
|
541 | PUTBACK; |
|
|
542 | call_sv ((SV *)GvCV (to_json), G_SCALAR); |
|
|
543 | SPAGAIN; |
|
|
544 | |
|
|
545 | // catch this surprisingly common error |
|
|
546 | if (SvROK (TOPs) && SvRV (TOPs) == sv) |
|
|
547 | croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv))); |
|
|
548 | |
|
|
549 | sv = POPs; |
|
|
550 | PUTBACK; |
|
|
551 | |
|
|
552 | encode_sv (enc, sv); |
|
|
553 | |
|
|
554 | FREETMPS; LEAVE; |
|
|
555 | } |
|
|
556 | else if (enc->json.flags & F_ALLOW_BLESSED) |
|
|
557 | encode_str (enc, "null", 4, 0); |
|
|
558 | else |
|
|
559 | croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it", |
|
|
560 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
|
|
561 | } |
|
|
562 | else if (enc->json.flags & F_ALLOW_BLESSED) |
|
|
563 | encode_str (enc, "null", 4, 0); |
|
|
564 | else |
|
|
565 | croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled", |
|
|
566 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
354 | } |
567 | } |
355 | } |
568 | } |
|
|
569 | else if (svt == SVt_PVHV) |
|
|
570 | encode_hv (enc, (HV *)sv); |
|
|
571 | else if (svt == SVt_PVAV) |
|
|
572 | encode_av (enc, (AV *)sv); |
|
|
573 | else if (svt < SVt_PVAV) |
|
|
574 | { |
|
|
575 | STRLEN len = 0; |
|
|
576 | char *pv = svt ? SvPV (sv, len) : 0; |
356 | |
577 | |
357 | --enc->indent; INDENT; encode_ch (enc, '}'); |
578 | if (len == 1 && *pv == '1') |
|
|
579 | encode_str (enc, "true", 4, 0); |
|
|
580 | else if (len == 1 && *pv == '0') |
|
|
581 | encode_str (enc, "false", 5, 0); |
|
|
582 | else |
|
|
583 | croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", |
|
|
584 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
|
|
585 | } |
|
|
586 | else |
|
|
587 | croak ("encountered %s, but JSON can only represent references to arrays or hashes", |
|
|
588 | SvPV_nolen (sv_2mortal (newRV_inc (sv)))); |
358 | } |
589 | } |
359 | |
590 | |
360 | static void |
591 | static void |
361 | encode_sv (enc_t *enc, SV *sv) |
592 | encode_sv (enc_t *enc, SV *sv) |
362 | { |
593 | { |
… | |
… | |
370 | encode_str (enc, str, len, SvUTF8 (sv)); |
601 | encode_str (enc, str, len, SvUTF8 (sv)); |
371 | encode_ch (enc, '"'); |
602 | encode_ch (enc, '"'); |
372 | } |
603 | } |
373 | else if (SvNOKp (sv)) |
604 | else if (SvNOKp (sv)) |
374 | { |
605 | { |
|
|
606 | // trust that perl will do the right thing w.r.t. JSON syntax. |
375 | need (enc, NV_DIG + 32); |
607 | need (enc, NV_DIG + 32); |
376 | Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); |
608 | Gconvert (SvNVX (sv), NV_DIG, 0, enc->cur); |
377 | enc->cur += strlen (enc->cur); |
609 | enc->cur += strlen (enc->cur); |
378 | } |
610 | } |
379 | else if (SvIOKp (sv)) |
611 | else if (SvIOKp (sv)) |
380 | { |
612 | { |
381 | need (enc, 64); |
613 | // we assume we can always read an IV as a UV |
|
|
614 | if (SvUV (sv) & ~(UV)0x7fff) |
|
|
615 | { |
|
|
616 | // large integer, use the (rather slow) snprintf way. |
|
|
617 | need (enc, sizeof (UV) * 3); |
382 | enc->cur += |
618 | enc->cur += |
383 | SvIsUV(sv) |
619 | SvIsUV(sv) |
384 | ? snprintf (enc->cur, 64, "%"UVuf, (UV)SvUVX (sv)) |
620 | ? snprintf (enc->cur, sizeof (UV) * 3, "%"UVuf, (UV)SvUVX (sv)) |
385 | : snprintf (enc->cur, 64, "%"IVdf, (IV)SvIVX (sv)); |
621 | : snprintf (enc->cur, sizeof (UV) * 3, "%"IVdf, (IV)SvIVX (sv)); |
|
|
622 | } |
|
|
623 | else |
|
|
624 | { |
|
|
625 | // optimise the "small number case" |
|
|
626 | // code will likely be branchless and use only a single multiplication |
|
|
627 | I32 i = SvIV (sv); |
|
|
628 | U32 u; |
|
|
629 | char digit, nz = 0; |
|
|
630 | |
|
|
631 | need (enc, 6); |
|
|
632 | |
|
|
633 | *enc->cur = '-'; enc->cur += i < 0 ? 1 : 0; |
|
|
634 | u = i < 0 ? -i : i; |
|
|
635 | |
|
|
636 | // convert to 4.28 fixed-point representation |
|
|
637 | u = u * ((0xfffffff + 10000) / 10000); // 10**5, 5 fractional digits |
|
|
638 | |
|
|
639 | // now output digit by digit, each time masking out the integer part |
|
|
640 | // and multiplying by 5 while moving the decimal point one to the right, |
|
|
641 | // resulting in a net multiplication by 10. |
|
|
642 | // we always write the digit to memory but conditionally increment |
|
|
643 | // the pointer, to ease the usage of conditional move instructions. |
|
|
644 | digit = u >> 28; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0xfffffff) * 5; |
|
|
645 | digit = u >> 27; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x7ffffff) * 5; |
|
|
646 | digit = u >> 26; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x3ffffff) * 5; |
|
|
647 | digit = u >> 25; *enc->cur = digit + '0'; enc->cur += (nz = nz || digit); u = (u & 0x1ffffff) * 5; |
|
|
648 | digit = u >> 24; *enc->cur = digit + '0'; enc->cur += 1; // correctly generate '0' |
|
|
649 | } |
386 | } |
650 | } |
387 | else if (SvROK (sv)) |
651 | else if (SvROK (sv)) |
388 | { |
652 | encode_rv (enc, SvRV (sv)); |
389 | if (!--enc->max_recurse) |
|
|
390 | croak ("data structure too deep (hit recursion limit)"); |
|
|
391 | |
|
|
392 | sv = SvRV (sv); |
|
|
393 | |
|
|
394 | switch (SvTYPE (sv)) |
|
|
395 | { |
|
|
396 | case SVt_PVAV: encode_av (enc, (AV *)sv); break; |
|
|
397 | case SVt_PVHV: encode_hv (enc, (HV *)sv); break; |
|
|
398 | |
|
|
399 | default: |
|
|
400 | croak ("JSON can only represent references to arrays or hashes"); |
|
|
401 | } |
|
|
402 | } |
|
|
403 | else if (!SvOK (sv)) |
653 | else if (!SvOK (sv)) |
404 | encode_str (enc, "null", 4, 0); |
654 | encode_str (enc, "null", 4, 0); |
405 | else |
655 | else |
406 | croak ("encountered perl type that JSON cannot handle"); |
656 | croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", |
|
|
657 | SvPV_nolen (sv), SvFLAGS (sv)); |
407 | } |
658 | } |
408 | |
659 | |
409 | static SV * |
660 | static SV * |
410 | encode_json (SV *scalar, UV flags) |
661 | encode_json (SV *scalar, JSON *json) |
411 | { |
662 | { |
412 | if (!(flags & F_ALLOW_NONREF) && !SvROK (scalar)) |
|
|
413 | croak ("hash- or arraref required (not a simple scalar, use allow_nonref to allow this)"); |
|
|
414 | |
|
|
415 | enc_t enc; |
663 | enc_t enc; |
416 | enc.flags = flags; |
664 | |
|
|
665 | if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar)) |
|
|
666 | croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)"); |
|
|
667 | |
|
|
668 | enc.json = *json; |
417 | enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); |
669 | enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); |
418 | enc.cur = SvPVX (enc.sv); |
670 | enc.cur = SvPVX (enc.sv); |
419 | enc.end = SvEND (enc.sv); |
671 | enc.end = SvEND (enc.sv); |
420 | enc.max_recurse = 0; |
|
|
421 | enc.indent = 0; |
672 | enc.indent = 0; |
|
|
673 | enc.maxdepth = DEC_DEPTH (enc.json.flags); |
422 | |
674 | |
423 | SvPOK_only (enc.sv); |
675 | SvPOK_only (enc.sv); |
424 | encode_sv (&enc, scalar); |
676 | encode_sv (&enc, scalar); |
425 | |
677 | |
|
|
678 | SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); |
|
|
679 | *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings |
|
|
680 | |
426 | if (!(flags & (F_ASCII | F_UTF8))) |
681 | if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8))) |
427 | SvUTF8_on (enc.sv); |
682 | SvUTF8_on (enc.sv); |
428 | |
683 | |
429 | SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); |
684 | if (enc.json.flags & F_SHRINK) |
|
|
685 | shrink (enc.sv); |
|
|
686 | |
430 | return enc.sv; |
687 | return enc.sv; |
431 | } |
688 | } |
432 | |
689 | |
433 | ///////////////////////////////////////////////////////////////////////////// |
690 | ///////////////////////////////////////////////////////////////////////////// |
|
|
691 | // decoder |
434 | |
692 | |
435 | #define WS \ |
693 | // structure used for decoding JSON |
|
|
694 | typedef struct |
|
|
695 | { |
|
|
696 | char *cur; // current parser pointer |
|
|
697 | char *end; // end of input string |
|
|
698 | const char *err; // parse error, if != 0 |
|
|
699 | JSON json; |
|
|
700 | U32 depth; // recursion depth |
|
|
701 | U32 maxdepth; // recursion depth limit |
|
|
702 | } dec_t; |
|
|
703 | |
|
|
704 | inline void |
|
|
705 | decode_ws (dec_t *dec) |
|
|
706 | { |
436 | for (;;) \ |
707 | for (;;) |
437 | { \ |
708 | { |
438 | char ch = *dec->cur; \ |
709 | char ch = *dec->cur; |
|
|
710 | |
439 | if (ch > 0x20 \ |
711 | if (ch > 0x20 |
440 | || (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) \ |
712 | || (ch != 0x20 && ch != 0x0a && ch != 0x0d && ch != 0x09)) |
441 | break; \ |
713 | break; |
|
|
714 | |
|
|
715 | if (ch == '#' && dec->json.flags & F_RELAXED) |
|
|
716 | ++dec->cur; |
|
|
717 | |
442 | ++dec->cur; \ |
718 | ++dec->cur; |
443 | } |
719 | } |
|
|
720 | } |
444 | |
721 | |
445 | #define ERR(reason) SB dec->err = reason; goto fail; SE |
722 | #define ERR(reason) SB dec->err = reason; goto fail; SE |
|
|
723 | |
446 | #define EXPECT_CH(ch) SB \ |
724 | #define EXPECT_CH(ch) SB \ |
447 | if (*dec->cur != ch) \ |
725 | if (*dec->cur != ch) \ |
448 | ERR (# ch " expected"); \ |
726 | ERR (# ch " expected"); \ |
449 | ++dec->cur; \ |
727 | ++dec->cur; \ |
450 | SE |
728 | SE |
451 | |
729 | |
|
|
730 | #define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)") |
|
|
731 | #define DEC_DEC_DEPTH --dec->depth |
|
|
732 | |
452 | static SV *decode_sv (dec_t *dec); |
733 | static SV *decode_sv (dec_t *dec); |
453 | |
734 | |
454 | static signed char decode_hexdigit[256]; |
735 | static signed char decode_hexdigit[256]; |
455 | |
736 | |
456 | static UV |
737 | static UV |
457 | decode_4hex (dec_t *dec) |
738 | decode_4hex (dec_t *dec) |
458 | { |
739 | { |
459 | signed char d1, d2, d3, d4; |
740 | signed char d1, d2, d3, d4; |
|
|
741 | unsigned char *cur = (unsigned char *)dec->cur; |
460 | |
742 | |
461 | d1 = decode_hexdigit [((unsigned char *)dec->cur) [0]]; |
743 | d1 = decode_hexdigit [cur [0]]; if (expect_false (d1 < 0)) ERR ("exactly four hexadecimal digits expected"); |
462 | if (d1 < 0) ERR ("four hexadecimal digits expected"); |
744 | d2 = decode_hexdigit [cur [1]]; if (expect_false (d2 < 0)) ERR ("exactly four hexadecimal digits expected"); |
463 | d2 = decode_hexdigit [((unsigned char *)dec->cur) [1]]; |
745 | d3 = decode_hexdigit [cur [2]]; if (expect_false (d3 < 0)) ERR ("exactly four hexadecimal digits expected"); |
464 | if (d2 < 0) ERR ("four hexadecimal digits expected"); |
746 | d4 = decode_hexdigit [cur [3]]; if (expect_false (d4 < 0)) ERR ("exactly four hexadecimal digits expected"); |
465 | d3 = decode_hexdigit [((unsigned char *)dec->cur) [2]]; |
|
|
466 | if (d3 < 0) ERR ("four hexadecimal digits expected"); |
|
|
467 | d4 = decode_hexdigit [((unsigned char *)dec->cur) [3]]; |
|
|
468 | if (d4 < 0) ERR ("four hexadecimal digits expected"); |
|
|
469 | |
747 | |
470 | dec->cur += 4; |
748 | dec->cur += 4; |
471 | |
749 | |
472 | return ((UV)d1) << 12 |
750 | return ((UV)d1) << 12 |
473 | | ((UV)d2) << 8 |
751 | | ((UV)d2) << 8 |
… | |
… | |
476 | |
754 | |
477 | fail: |
755 | fail: |
478 | return (UV)-1; |
756 | return (UV)-1; |
479 | } |
757 | } |
480 | |
758 | |
481 | #define APPEND_GROW(n) SB \ |
|
|
482 | if (cur + (n) >= end) \ |
|
|
483 | { \ |
|
|
484 | STRLEN ofs = cur - SvPVX (sv); \ |
|
|
485 | SvGROW (sv, ofs + (n) + 1); \ |
|
|
486 | cur = SvPVX (sv) + ofs; \ |
|
|
487 | end = SvEND (sv); \ |
|
|
488 | } \ |
|
|
489 | SE |
|
|
490 | |
|
|
491 | #define APPEND_CH(ch) SB \ |
|
|
492 | APPEND_GROW (1); \ |
|
|
493 | *cur++ = (ch); \ |
|
|
494 | SE |
|
|
495 | |
|
|
496 | static SV * |
759 | static SV * |
497 | decode_str (dec_t *dec) |
760 | decode_str (dec_t *dec) |
498 | { |
761 | { |
499 | SV *sv = NEWSV (0,2); |
762 | SV *sv = 0; |
500 | int utf8 = 0; |
763 | int utf8 = 0; |
501 | char *cur = SvPVX (sv); |
764 | char *dec_cur = dec->cur; |
502 | char *end = SvEND (sv); |
|
|
503 | |
765 | |
504 | for (;;) |
766 | do |
505 | { |
767 | { |
506 | unsigned char ch = *(unsigned char *)dec->cur; |
768 | char buf [SHORT_STRING_LEN + UTF8_MAXBYTES]; |
|
|
769 | char *cur = buf; |
507 | |
770 | |
508 | if (ch == '"') |
771 | do |
509 | break; |
|
|
510 | else if (ch == '\\') |
|
|
511 | { |
772 | { |
512 | switch (*++dec->cur) |
773 | unsigned char ch = *(unsigned char *)dec_cur++; |
|
|
774 | |
|
|
775 | if (expect_false (ch == '"')) |
|
|
776 | { |
|
|
777 | --dec_cur; |
|
|
778 | break; |
513 | { |
779 | } |
514 | case '\\': |
780 | else if (expect_false (ch == '\\')) |
515 | case '/': |
781 | { |
516 | case '"': APPEND_CH (*dec->cur++); break; |
782 | switch (*dec_cur) |
517 | |
|
|
518 | case 'b': APPEND_CH ('\010'); ++dec->cur; break; |
|
|
519 | case 't': APPEND_CH ('\011'); ++dec->cur; break; |
|
|
520 | case 'n': APPEND_CH ('\012'); ++dec->cur; break; |
|
|
521 | case 'f': APPEND_CH ('\014'); ++dec->cur; break; |
|
|
522 | case 'r': APPEND_CH ('\015'); ++dec->cur; break; |
|
|
523 | |
|
|
524 | case 'u': |
|
|
525 | { |
783 | { |
526 | UV lo, hi; |
784 | case '\\': |
527 | ++dec->cur; |
785 | case '/': |
|
|
786 | case '"': *cur++ = *dec_cur++; break; |
528 | |
787 | |
529 | hi = decode_4hex (dec); |
788 | case 'b': ++dec_cur; *cur++ = '\010'; break; |
530 | if (hi == (UV)-1) |
789 | case 't': ++dec_cur; *cur++ = '\011'; break; |
531 | goto fail; |
790 | case 'n': ++dec_cur; *cur++ = '\012'; break; |
|
|
791 | case 'f': ++dec_cur; *cur++ = '\014'; break; |
|
|
792 | case 'r': ++dec_cur; *cur++ = '\015'; break; |
532 | |
793 | |
533 | // possibly a surrogate pair |
794 | case 'u': |
534 | if (hi >= 0xd800 && hi < 0xdc00) |
|
|
535 | { |
795 | { |
536 | if (dec->cur [0] != '\\' || dec->cur [1] != 'u') |
796 | UV lo, hi; |
537 | ERR ("missing low surrogate character in surrogate pair"); |
797 | ++dec_cur; |
538 | |
798 | |
539 | dec->cur += 2; |
799 | dec->cur = dec_cur; |
540 | |
|
|
541 | lo = decode_4hex (dec); |
800 | hi = decode_4hex (dec); |
|
|
801 | dec_cur = dec->cur; |
542 | if (lo == (UV)-1) |
802 | if (hi == (UV)-1) |
543 | goto fail; |
803 | goto fail; |
544 | |
804 | |
|
|
805 | // possibly a surrogate pair |
|
|
806 | if (hi >= 0xd800) |
|
|
807 | if (hi < 0xdc00) |
|
|
808 | { |
|
|
809 | if (dec_cur [0] != '\\' || dec_cur [1] != 'u') |
|
|
810 | ERR ("missing low surrogate character in surrogate pair"); |
|
|
811 | |
|
|
812 | dec_cur += 2; |
|
|
813 | |
|
|
814 | dec->cur = dec_cur; |
|
|
815 | lo = decode_4hex (dec); |
|
|
816 | dec_cur = dec->cur; |
|
|
817 | if (lo == (UV)-1) |
|
|
818 | goto fail; |
|
|
819 | |
545 | if (lo < 0xdc00 || lo >= 0xe000) |
820 | if (lo < 0xdc00 || lo >= 0xe000) |
546 | ERR ("surrogate pair expected"); |
821 | ERR ("surrogate pair expected"); |
547 | |
822 | |
548 | hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000; |
823 | hi = (hi - 0xD800) * 0x400 + (lo - 0xDC00) + 0x10000; |
|
|
824 | } |
|
|
825 | else if (hi < 0xe000) |
|
|
826 | ERR ("missing high surrogate character in surrogate pair"); |
|
|
827 | |
|
|
828 | if (hi >= 0x80) |
|
|
829 | { |
|
|
830 | utf8 = 1; |
|
|
831 | |
|
|
832 | cur = (char *)uvuni_to_utf8_flags (cur, hi, 0); |
|
|
833 | } |
|
|
834 | else |
|
|
835 | *cur++ = hi; |
549 | } |
836 | } |
550 | else if (hi >= 0xdc00 && hi < 0xe000) |
|
|
551 | ERR ("missing high surrogate character in surrogate pair"); |
|
|
552 | |
|
|
553 | if (hi >= 0x80) |
|
|
554 | { |
837 | break; |
555 | utf8 = 1; |
|
|
556 | |
838 | |
557 | APPEND_GROW (4); // at most 4 bytes for 21 bits |
|
|
558 | cur = (char *)uvuni_to_utf8_flags (cur, hi, 0); |
|
|
559 | } |
|
|
560 | else |
839 | default: |
561 | APPEND_CH (hi); |
840 | --dec_cur; |
|
|
841 | ERR ("illegal backslash escape sequence in string"); |
562 | } |
842 | } |
563 | break; |
843 | } |
|
|
844 | else if (expect_true (ch >= 0x20 && ch <= 0x7f)) |
|
|
845 | *cur++ = ch; |
|
|
846 | else if (ch >= 0x80) |
|
|
847 | { |
|
|
848 | STRLEN clen; |
|
|
849 | UV uch; |
564 | |
850 | |
565 | default: |
|
|
566 | --dec->cur; |
851 | --dec_cur; |
567 | ERR ("illegal backslash escape sequence in string"); |
852 | |
|
|
853 | uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen); |
|
|
854 | if (clen == (STRLEN)-1) |
|
|
855 | ERR ("malformed UTF-8 character in JSON string"); |
|
|
856 | |
|
|
857 | do |
|
|
858 | *cur++ = *dec_cur++; |
|
|
859 | while (--clen); |
|
|
860 | |
|
|
861 | utf8 = 1; |
|
|
862 | } |
|
|
863 | else |
|
|
864 | { |
|
|
865 | --dec_cur; |
|
|
866 | |
|
|
867 | if (!ch) |
|
|
868 | ERR ("unexpected end of string while parsing JSON string"); |
|
|
869 | else |
|
|
870 | ERR ("invalid character encountered while parsing JSON string"); |
568 | } |
871 | } |
569 | } |
872 | } |
570 | else if (ch >= 0x20 && ch <= 0x7f) |
873 | while (cur < buf + SHORT_STRING_LEN); |
571 | APPEND_CH (*dec->cur++); |
874 | |
572 | else if (ch >= 0x80) |
|
|
573 | { |
875 | { |
574 | STRLEN clen; |
876 | STRLEN len = cur - buf; |
575 | UV uch = utf8n_to_uvuni (dec->cur, dec->end - dec->cur, &clen, UTF8_CHECK_ONLY); |
|
|
576 | if (clen == (STRLEN)-1) |
|
|
577 | ERR ("malformed UTF-8 character in string, cannot convert to JSON"); |
|
|
578 | |
877 | |
579 | APPEND_GROW (clen); |
878 | if (sv) |
580 | do |
|
|
581 | { |
879 | { |
582 | *cur++ = *dec->cur++; |
880 | SvGROW (sv, SvCUR (sv) + len + 1); |
|
|
881 | memcpy (SvPVX (sv) + SvCUR (sv), buf, len); |
|
|
882 | SvCUR_set (sv, SvCUR (sv) + len); |
583 | } |
883 | } |
584 | while (--clen); |
|
|
585 | |
|
|
586 | utf8 = 1; |
|
|
587 | } |
|
|
588 | else if (dec->cur == dec->end) |
|
|
589 | ERR ("unexpected end of string while parsing json string"); |
|
|
590 | else |
884 | else |
591 | ERR ("invalid character encountered"); |
885 | sv = newSVpvn (buf, len); |
592 | } |
886 | } |
|
|
887 | } |
|
|
888 | while (*dec_cur != '"'); |
593 | |
889 | |
594 | ++dec->cur; |
890 | ++dec_cur; |
595 | |
891 | |
596 | SvCUR_set (sv, cur - SvPVX (sv)); |
892 | if (sv) |
597 | |
893 | { |
598 | SvPOK_only (sv); |
894 | SvPOK_only (sv); |
599 | *SvEND (sv) = 0; |
895 | *SvEND (sv) = 0; |
600 | |
896 | |
601 | if (utf8) |
897 | if (utf8) |
602 | SvUTF8_on (sv); |
898 | SvUTF8_on (sv); |
|
|
899 | } |
|
|
900 | else |
|
|
901 | sv = newSVpvn ("", 0); |
603 | |
902 | |
|
|
903 | dec->cur = dec_cur; |
604 | return sv; |
904 | return sv; |
605 | |
905 | |
606 | fail: |
906 | fail: |
607 | SvREFCNT_dec (sv); |
907 | dec->cur = dec_cur; |
608 | return 0; |
908 | return 0; |
609 | } |
909 | } |
610 | |
910 | |
611 | static SV * |
911 | static SV * |
612 | decode_num (dec_t *dec) |
912 | decode_num (dec_t *dec) |
… | |
… | |
670 | is_nv = 1; |
970 | is_nv = 1; |
671 | } |
971 | } |
672 | |
972 | |
673 | if (!is_nv) |
973 | if (!is_nv) |
674 | { |
974 | { |
675 | UV uv; |
975 | int len = dec->cur - start; |
676 | int numtype = grok_number (start, dec->cur - start, &uv); |
976 | |
677 | if (numtype & IS_NUMBER_IN_UV) |
977 | // special case the rather common 1..4-digit-int case, assumes 32 bit ints or so |
678 | if (numtype & IS_NUMBER_NEG) |
978 | if (*start == '-') |
|
|
979 | switch (len) |
679 | { |
980 | { |
680 | if (uv < (UV)IV_MIN) |
981 | case 2: return newSViv (-( start [1] - '0' * 1)); |
681 | return newSViv (-(IV)uv); |
982 | case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11)); |
|
|
983 | case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111)); |
|
|
984 | case 5: return newSViv (-(start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111)); |
682 | } |
985 | } |
|
|
986 | else |
|
|
987 | switch (len) |
|
|
988 | { |
|
|
989 | case 1: return newSViv ( start [0] - '0' * 1); |
|
|
990 | case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); |
|
|
991 | case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111); |
|
|
992 | case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111); |
|
|
993 | } |
|
|
994 | |
|
|
995 | { |
|
|
996 | UV uv; |
|
|
997 | int numtype = grok_number (start, len, &uv); |
|
|
998 | if (numtype & IS_NUMBER_IN_UV) |
|
|
999 | if (numtype & IS_NUMBER_NEG) |
|
|
1000 | { |
|
|
1001 | if (uv < (UV)IV_MIN) |
|
|
1002 | return newSViv (-(IV)uv); |
|
|
1003 | } |
683 | else |
1004 | else |
684 | return newSVuv (uv); |
1005 | return newSVuv (uv); |
685 | } |
1006 | } |
686 | |
1007 | |
|
|
1008 | len -= *start == '-' ? 1 : 0; |
|
|
1009 | |
|
|
1010 | // does not fit into IV or UV, try NV |
|
|
1011 | if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len) |
|
|
1012 | #if defined (LDBL_DIG) |
|
|
1013 | || (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len) |
|
|
1014 | #endif |
|
|
1015 | ) |
|
|
1016 | // fits into NV without loss of precision |
|
|
1017 | return newSVnv (Atof (start)); |
|
|
1018 | |
|
|
1019 | // everything else fails, convert it to a string |
|
|
1020 | return newSVpvn (start, dec->cur - start); |
|
|
1021 | } |
|
|
1022 | |
|
|
1023 | // loss of precision here |
687 | return newSVnv (Atof (start)); |
1024 | return newSVnv (Atof (start)); |
688 | |
1025 | |
689 | fail: |
1026 | fail: |
690 | return 0; |
1027 | return 0; |
691 | } |
1028 | } |
… | |
… | |
693 | static SV * |
1030 | static SV * |
694 | decode_av (dec_t *dec) |
1031 | decode_av (dec_t *dec) |
695 | { |
1032 | { |
696 | AV *av = newAV (); |
1033 | AV *av = newAV (); |
697 | |
1034 | |
698 | WS; |
1035 | DEC_INC_DEPTH; |
|
|
1036 | decode_ws (dec); |
|
|
1037 | |
699 | if (*dec->cur == ']') |
1038 | if (*dec->cur == ']') |
700 | ++dec->cur; |
1039 | ++dec->cur; |
701 | else |
1040 | else |
702 | for (;;) |
1041 | for (;;) |
703 | { |
1042 | { |
… | |
… | |
707 | if (!value) |
1046 | if (!value) |
708 | goto fail; |
1047 | goto fail; |
709 | |
1048 | |
710 | av_push (av, value); |
1049 | av_push (av, value); |
711 | |
1050 | |
712 | WS; |
1051 | decode_ws (dec); |
713 | |
1052 | |
714 | if (*dec->cur == ']') |
1053 | if (*dec->cur == ']') |
715 | { |
1054 | { |
716 | ++dec->cur; |
1055 | ++dec->cur; |
717 | break; |
1056 | break; |
… | |
… | |
719 | |
1058 | |
720 | if (*dec->cur != ',') |
1059 | if (*dec->cur != ',') |
721 | ERR (", or ] expected while parsing array"); |
1060 | ERR (", or ] expected while parsing array"); |
722 | |
1061 | |
723 | ++dec->cur; |
1062 | ++dec->cur; |
|
|
1063 | |
|
|
1064 | decode_ws (dec); |
|
|
1065 | |
|
|
1066 | if (*dec->cur == ']' && dec->json.flags & F_RELAXED) |
|
|
1067 | { |
|
|
1068 | ++dec->cur; |
|
|
1069 | break; |
|
|
1070 | } |
724 | } |
1071 | } |
725 | |
1072 | |
|
|
1073 | DEC_DEC_DEPTH; |
726 | return newRV_noinc ((SV *)av); |
1074 | return newRV_noinc ((SV *)av); |
727 | |
1075 | |
728 | fail: |
1076 | fail: |
729 | SvREFCNT_dec (av); |
1077 | SvREFCNT_dec (av); |
|
|
1078 | DEC_DEC_DEPTH; |
730 | return 0; |
1079 | return 0; |
731 | } |
1080 | } |
732 | |
1081 | |
733 | static SV * |
1082 | static SV * |
734 | decode_hv (dec_t *dec) |
1083 | decode_hv (dec_t *dec) |
735 | { |
1084 | { |
|
|
1085 | SV *sv; |
736 | HV *hv = newHV (); |
1086 | HV *hv = newHV (); |
737 | |
1087 | |
738 | WS; |
1088 | DEC_INC_DEPTH; |
|
|
1089 | decode_ws (dec); |
|
|
1090 | |
739 | if (*dec->cur == '}') |
1091 | if (*dec->cur == '}') |
740 | ++dec->cur; |
1092 | ++dec->cur; |
741 | else |
1093 | else |
742 | for (;;) |
1094 | for (;;) |
743 | { |
1095 | { |
744 | SV *key, *value; |
|
|
745 | |
|
|
746 | WS; EXPECT_CH ('"'); |
1096 | EXPECT_CH ('"'); |
747 | |
1097 | |
748 | key = decode_str (dec); |
1098 | // heuristic: assume that |
749 | if (!key) |
1099 | // a) decode_str + hv_store_ent are abysmally slow. |
750 | goto fail; |
1100 | // b) most hash keys are short, simple ascii text. |
|
|
1101 | // => try to "fast-match" such strings to avoid |
|
|
1102 | // the overhead of decode_str + hv_store_ent. |
|
|
1103 | { |
|
|
1104 | SV *value; |
|
|
1105 | char *p = dec->cur; |
|
|
1106 | char *e = p + 24; // only try up to 24 bytes |
751 | |
1107 | |
752 | WS; EXPECT_CH (':'); |
1108 | for (;;) |
753 | |
|
|
754 | value = decode_sv (dec); |
|
|
755 | if (!value) |
|
|
756 | { |
1109 | { |
|
|
1110 | // the >= 0x80 is true on most architectures |
|
|
1111 | if (p == e || *p < 0x20 || *p >= 0x80 || *p == '\\') |
|
|
1112 | { |
|
|
1113 | // slow path, back up and use decode_str |
|
|
1114 | SV *key = decode_str (dec); |
|
|
1115 | if (!key) |
|
|
1116 | goto fail; |
|
|
1117 | |
|
|
1118 | decode_ws (dec); EXPECT_CH (':'); |
|
|
1119 | |
|
|
1120 | decode_ws (dec); |
|
|
1121 | value = decode_sv (dec); |
|
|
1122 | if (!value) |
|
|
1123 | { |
|
|
1124 | SvREFCNT_dec (key); |
|
|
1125 | goto fail; |
|
|
1126 | } |
|
|
1127 | |
|
|
1128 | hv_store_ent (hv, key, value, 0); |
757 | SvREFCNT_dec (key); |
1129 | SvREFCNT_dec (key); |
|
|
1130 | |
|
|
1131 | break; |
|
|
1132 | } |
|
|
1133 | else if (*p == '"') |
|
|
1134 | { |
|
|
1135 | // fast path, got a simple key |
|
|
1136 | char *key = dec->cur; |
|
|
1137 | int len = p - key; |
|
|
1138 | dec->cur = p + 1; |
|
|
1139 | |
|
|
1140 | decode_ws (dec); EXPECT_CH (':'); |
|
|
1141 | |
|
|
1142 | decode_ws (dec); |
|
|
1143 | value = decode_sv (dec); |
|
|
1144 | if (!value) |
758 | goto fail; |
1145 | goto fail; |
|
|
1146 | |
|
|
1147 | hv_store (hv, key, len, value, 0); |
|
|
1148 | |
|
|
1149 | break; |
|
|
1150 | } |
|
|
1151 | |
|
|
1152 | ++p; |
759 | } |
1153 | } |
760 | |
|
|
761 | //TODO: optimise |
|
|
762 | hv_store_ent (hv, key, value, 0); |
|
|
763 | |
|
|
764 | WS; |
1154 | } |
|
|
1155 | |
|
|
1156 | decode_ws (dec); |
765 | |
1157 | |
766 | if (*dec->cur == '}') |
1158 | if (*dec->cur == '}') |
767 | { |
1159 | { |
768 | ++dec->cur; |
1160 | ++dec->cur; |
769 | break; |
1161 | break; |
… | |
… | |
771 | |
1163 | |
772 | if (*dec->cur != ',') |
1164 | if (*dec->cur != ',') |
773 | ERR (", or } expected while parsing object/hash"); |
1165 | ERR (", or } expected while parsing object/hash"); |
774 | |
1166 | |
775 | ++dec->cur; |
1167 | ++dec->cur; |
|
|
1168 | |
|
|
1169 | decode_ws (dec); |
|
|
1170 | |
|
|
1171 | if (*dec->cur == '}' && dec->json.flags & F_RELAXED) |
|
|
1172 | { |
|
|
1173 | ++dec->cur; |
|
|
1174 | break; |
|
|
1175 | } |
776 | } |
1176 | } |
777 | |
1177 | |
|
|
1178 | DEC_DEC_DEPTH; |
778 | return newRV_noinc ((SV *)hv); |
1179 | sv = newRV_noinc ((SV *)hv); |
|
|
1180 | |
|
|
1181 | // check filter callbacks |
|
|
1182 | if (dec->json.flags & F_HOOK) |
|
|
1183 | { |
|
|
1184 | if (dec->json.cb_sk_object && HvKEYS (hv) == 1) |
|
|
1185 | { |
|
|
1186 | HE *cb, *he; |
|
|
1187 | |
|
|
1188 | hv_iterinit (hv); |
|
|
1189 | he = hv_iternext (hv); |
|
|
1190 | hv_iterinit (hv); |
|
|
1191 | |
|
|
1192 | // the next line creates a mortal sv each time its called. |
|
|
1193 | // might want to optimise this for common cases. |
|
|
1194 | cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0); |
|
|
1195 | |
|
|
1196 | if (cb) |
|
|
1197 | { |
|
|
1198 | dSP; |
|
|
1199 | int count; |
|
|
1200 | |
|
|
1201 | ENTER; SAVETMPS; PUSHMARK (SP); |
|
|
1202 | XPUSHs (HeVAL (he)); |
|
|
1203 | |
|
|
1204 | PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN; |
|
|
1205 | |
|
|
1206 | if (count == 1) |
|
|
1207 | { |
|
|
1208 | sv = newSVsv (POPs); |
|
|
1209 | FREETMPS; LEAVE; |
|
|
1210 | return sv; |
|
|
1211 | } |
|
|
1212 | |
|
|
1213 | FREETMPS; LEAVE; |
|
|
1214 | } |
|
|
1215 | } |
|
|
1216 | |
|
|
1217 | if (dec->json.cb_object) |
|
|
1218 | { |
|
|
1219 | dSP; |
|
|
1220 | int count; |
|
|
1221 | |
|
|
1222 | ENTER; SAVETMPS; PUSHMARK (SP); |
|
|
1223 | XPUSHs (sv_2mortal (sv)); |
|
|
1224 | |
|
|
1225 | PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN; |
|
|
1226 | |
|
|
1227 | if (count == 1) |
|
|
1228 | { |
|
|
1229 | sv = newSVsv (POPs); |
|
|
1230 | FREETMPS; LEAVE; |
|
|
1231 | return sv; |
|
|
1232 | } |
|
|
1233 | |
|
|
1234 | SvREFCNT_inc (sv); |
|
|
1235 | FREETMPS; LEAVE; |
|
|
1236 | } |
|
|
1237 | } |
|
|
1238 | |
|
|
1239 | return sv; |
779 | |
1240 | |
780 | fail: |
1241 | fail: |
781 | SvREFCNT_dec (hv); |
1242 | SvREFCNT_dec (hv); |
|
|
1243 | DEC_DEC_DEPTH; |
782 | return 0; |
1244 | return 0; |
783 | } |
1245 | } |
784 | |
1246 | |
785 | static SV * |
1247 | static SV * |
786 | decode_sv (dec_t *dec) |
1248 | decode_sv (dec_t *dec) |
787 | { |
1249 | { |
788 | WS; |
1250 | // the beauty of JSON: you need exactly one character lookahead |
|
|
1251 | // to parse anything. |
789 | switch (*dec->cur) |
1252 | switch (*dec->cur) |
790 | { |
1253 | { |
791 | case '"': ++dec->cur; return decode_str (dec); |
1254 | case '"': ++dec->cur; return decode_str (dec); |
792 | case '[': ++dec->cur; return decode_av (dec); |
1255 | case '[': ++dec->cur; return decode_av (dec); |
793 | case '{': ++dec->cur; return decode_hv (dec); |
1256 | case '{': ++dec->cur; return decode_hv (dec); |
… | |
… | |
799 | |
1262 | |
800 | case 't': |
1263 | case 't': |
801 | if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) |
1264 | if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) |
802 | { |
1265 | { |
803 | dec->cur += 4; |
1266 | dec->cur += 4; |
804 | return newSViv (1); |
1267 | #if JSON_SLOW |
|
|
1268 | json_true = get_sv ("JSON::XS::true", 1); SvREADONLY_on (json_true); |
|
|
1269 | #endif |
|
|
1270 | return SvREFCNT_inc (json_true); |
805 | } |
1271 | } |
806 | else |
1272 | else |
807 | ERR ("'true' expected"); |
1273 | ERR ("'true' expected"); |
808 | |
1274 | |
809 | break; |
1275 | break; |
810 | |
1276 | |
811 | case 'f': |
1277 | case 'f': |
812 | if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) |
1278 | if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5)) |
813 | { |
1279 | { |
814 | dec->cur += 5; |
1280 | dec->cur += 5; |
815 | return newSViv (0); |
1281 | #if JSON_SLOW |
|
|
1282 | json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false); |
|
|
1283 | #endif |
|
|
1284 | return SvREFCNT_inc (json_false); |
816 | } |
1285 | } |
817 | else |
1286 | else |
818 | ERR ("'false' expected"); |
1287 | ERR ("'false' expected"); |
819 | |
1288 | |
820 | break; |
1289 | break; |
… | |
… | |
829 | ERR ("'null' expected"); |
1298 | ERR ("'null' expected"); |
830 | |
1299 | |
831 | break; |
1300 | break; |
832 | |
1301 | |
833 | default: |
1302 | default: |
834 | ERR ("malformed json string"); |
1303 | ERR ("malformed JSON string, neither array, object, number, string or atom"); |
835 | break; |
1304 | break; |
836 | } |
1305 | } |
837 | |
1306 | |
838 | fail: |
1307 | fail: |
839 | return 0; |
1308 | return 0; |
840 | } |
1309 | } |
841 | |
1310 | |
842 | static SV * |
1311 | static SV * |
843 | decode_json (SV *string, UV flags) |
1312 | decode_json (SV *string, JSON *json, UV *offset_return) |
844 | { |
1313 | { |
|
|
1314 | dec_t dec; |
|
|
1315 | UV offset; |
845 | SV *sv; |
1316 | SV *sv; |
846 | |
1317 | |
|
|
1318 | SvGETMAGIC (string); |
|
|
1319 | SvUPGRADE (string, SVt_PV); |
|
|
1320 | |
|
|
1321 | if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags)) |
|
|
1322 | croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", |
|
|
1323 | (unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags)); |
|
|
1324 | |
847 | if (flags & F_UTF8) |
1325 | if (json->flags & F_UTF8) |
848 | sv_utf8_downgrade (string, 0); |
1326 | sv_utf8_downgrade (string, 0); |
849 | else |
1327 | else |
850 | sv_utf8_upgrade (string); |
1328 | sv_utf8_upgrade (string); |
851 | |
1329 | |
852 | SvGROW (string, SvCUR (string) + 1); // should basically be a NOP |
1330 | SvGROW (string, SvCUR (string) + 1); // should basically be a NOP |
853 | |
1331 | |
854 | dec_t dec; |
1332 | dec.json = *json; |
855 | dec.flags = flags; |
|
|
856 | dec.cur = SvPVX (string); |
1333 | dec.cur = SvPVX (string); |
857 | dec.end = SvEND (string); |
1334 | dec.end = SvEND (string); |
858 | dec.err = 0; |
1335 | dec.err = 0; |
|
|
1336 | dec.depth = 0; |
|
|
1337 | dec.maxdepth = DEC_DEPTH (dec.json.flags); |
859 | |
1338 | |
|
|
1339 | if (dec.json.cb_object || dec.json.cb_sk_object) |
|
|
1340 | dec.json.flags |= F_HOOK; |
|
|
1341 | |
|
|
1342 | *dec.end = 0; // this should basically be a nop, too, but make sure it's there |
|
|
1343 | |
|
|
1344 | decode_ws (&dec); |
860 | sv = decode_sv (&dec); |
1345 | sv = decode_sv (&dec); |
861 | |
1346 | |
|
|
1347 | if (!(offset_return || !sv)) |
|
|
1348 | { |
|
|
1349 | // check for trailing garbage |
|
|
1350 | decode_ws (&dec); |
|
|
1351 | |
|
|
1352 | if (*dec.cur) |
|
|
1353 | { |
|
|
1354 | dec.err = "garbage after JSON object"; |
|
|
1355 | SvREFCNT_dec (sv); |
|
|
1356 | sv = 0; |
|
|
1357 | } |
|
|
1358 | } |
|
|
1359 | |
|
|
1360 | if (offset_return || !sv) |
|
|
1361 | { |
|
|
1362 | offset = dec.json.flags & F_UTF8 |
|
|
1363 | ? dec.cur - SvPVX (string) |
|
|
1364 | : utf8_distance (dec.cur, SvPVX (string)); |
|
|
1365 | |
|
|
1366 | if (offset_return) |
|
|
1367 | *offset_return = offset; |
|
|
1368 | } |
|
|
1369 | |
862 | if (!sv) |
1370 | if (!sv) |
863 | { |
1371 | { |
864 | IV offset = utf8_distance (dec.cur, SvPVX (string)); |
|
|
865 | SV *uni = sv_newmortal (); |
1372 | SV *uni = sv_newmortal (); |
|
|
1373 | |
866 | // horrible hack to silence warning inside pv_uni_display |
1374 | // horrible hack to silence warning inside pv_uni_display |
867 | COP cop; |
1375 | COP cop = *PL_curcop; |
868 | memset (&cop, 0, sizeof (cop)); |
|
|
869 | cop.cop_warnings = pWARN_NONE; |
1376 | cop.cop_warnings = pWARN_NONE; |
|
|
1377 | ENTER; |
870 | SAVEVPTR (PL_curcop); |
1378 | SAVEVPTR (PL_curcop); |
871 | PL_curcop = &cop; |
1379 | PL_curcop = &cop; |
872 | |
|
|
873 | pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); |
1380 | pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); |
|
|
1381 | LEAVE; |
|
|
1382 | |
874 | croak ("%s, at character offset %d (%s)", |
1383 | croak ("%s, at character offset %d [\"%s\"]", |
875 | dec.err, |
1384 | dec.err, |
876 | (int)offset, |
1385 | (int)offset, |
877 | dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); |
1386 | dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); |
878 | } |
1387 | } |
879 | |
1388 | |
880 | sv = sv_2mortal (sv); |
1389 | sv = sv_2mortal (sv); |
881 | |
1390 | |
882 | if (!(dec.flags & F_ALLOW_NONREF) && !SvROK (sv)) |
1391 | if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv)) |
883 | croak ("JSON object or array expected (but number, string, true, false or null found, use allow_nonref to allow this)"); |
1392 | croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)"); |
884 | |
1393 | |
885 | return sv; |
1394 | return sv; |
886 | } |
1395 | } |
887 | |
1396 | |
|
|
1397 | ///////////////////////////////////////////////////////////////////////////// |
|
|
1398 | // XS interface functions |
|
|
1399 | |
888 | MODULE = JSON::XS PACKAGE = JSON::XS |
1400 | MODULE = JSON::XS PACKAGE = JSON::XS |
889 | |
1401 | |
890 | BOOT: |
1402 | BOOT: |
891 | { |
1403 | { |
892 | int i; |
1404 | int i; |
893 | |
1405 | |
894 | memset (decode_hexdigit, 0xff, 256); |
|
|
895 | for (i = 10; i--; ) |
1406 | for (i = 0; i < 256; ++i) |
896 | decode_hexdigit ['0' + i] = i; |
1407 | decode_hexdigit [i] = |
|
|
1408 | i >= '0' && i <= '9' ? i - '0' |
|
|
1409 | : i >= 'a' && i <= 'f' ? i - 'a' + 10 |
|
|
1410 | : i >= 'A' && i <= 'F' ? i - 'A' + 10 |
|
|
1411 | : -1; |
897 | |
1412 | |
898 | for (i = 7; i--; ) |
1413 | json_stash = gv_stashpv ("JSON::XS" , 1); |
|
|
1414 | json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1); |
|
|
1415 | |
|
|
1416 | json_true = get_sv ("JSON::XS::true" , 1); SvREADONLY_on (json_true ); |
|
|
1417 | json_false = get_sv ("JSON::XS::false", 1); SvREADONLY_on (json_false); |
|
|
1418 | } |
|
|
1419 | |
|
|
1420 | PROTOTYPES: DISABLE |
|
|
1421 | |
|
|
1422 | void CLONE (...) |
|
|
1423 | CODE: |
|
|
1424 | json_stash = 0; |
|
|
1425 | json_boolean_stash = 0; |
|
|
1426 | |
|
|
1427 | void new (char *klass) |
|
|
1428 | PPCODE: |
|
|
1429 | { |
|
|
1430 | SV *pv = NEWSV (0, sizeof (JSON)); |
|
|
1431 | SvPOK_only (pv); |
|
|
1432 | Zero (SvPVX (pv), 1, JSON); |
|
|
1433 | ((JSON *)SvPVX (pv))->flags = F_DEFAULT; |
|
|
1434 | XPUSHs (sv_2mortal (sv_bless (newRV_noinc (pv), JSON_STASH))); |
|
|
1435 | } |
|
|
1436 | |
|
|
1437 | void ascii (JSON *self, int enable = 1) |
|
|
1438 | ALIAS: |
|
|
1439 | ascii = F_ASCII |
|
|
1440 | latin1 = F_LATIN1 |
|
|
1441 | utf8 = F_UTF8 |
|
|
1442 | indent = F_INDENT |
|
|
1443 | canonical = F_CANONICAL |
|
|
1444 | space_before = F_SPACE_BEFORE |
|
|
1445 | space_after = F_SPACE_AFTER |
|
|
1446 | pretty = F_PRETTY |
|
|
1447 | allow_nonref = F_ALLOW_NONREF |
|
|
1448 | shrink = F_SHRINK |
|
|
1449 | allow_blessed = F_ALLOW_BLESSED |
|
|
1450 | convert_blessed = F_CONV_BLESSED |
|
|
1451 | relaxed = F_RELAXED |
|
|
1452 | PPCODE: |
|
|
1453 | { |
|
|
1454 | if (enable) |
|
|
1455 | self->flags |= ix; |
|
|
1456 | else |
|
|
1457 | self->flags &= ~ix; |
|
|
1458 | |
|
|
1459 | XPUSHs (ST (0)); |
|
|
1460 | } |
|
|
1461 | |
|
|
1462 | void max_depth (JSON *self, UV max_depth = 0x80000000UL) |
|
|
1463 | PPCODE: |
|
|
1464 | { |
|
|
1465 | UV log2 = 0; |
|
|
1466 | |
|
|
1467 | if (max_depth > 0x80000000UL) max_depth = 0x80000000UL; |
|
|
1468 | |
|
|
1469 | while ((1UL << log2) < max_depth) |
|
|
1470 | ++log2; |
|
|
1471 | |
|
|
1472 | self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH); |
|
|
1473 | |
|
|
1474 | XPUSHs (ST (0)); |
|
|
1475 | } |
|
|
1476 | |
|
|
1477 | void max_size (JSON *self, UV max_size = 0) |
|
|
1478 | PPCODE: |
|
|
1479 | { |
|
|
1480 | UV log2 = 0; |
|
|
1481 | |
|
|
1482 | if (max_size > 0x80000000UL) max_size = 0x80000000UL; |
|
|
1483 | if (max_size == 1) max_size = 2; |
|
|
1484 | |
|
|
1485 | while ((1UL << log2) < max_size) |
|
|
1486 | ++log2; |
|
|
1487 | |
|
|
1488 | self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE); |
|
|
1489 | |
|
|
1490 | XPUSHs (ST (0)); |
|
|
1491 | } |
|
|
1492 | |
|
|
1493 | void filter_json_object (JSON *self, SV *cb = &PL_sv_undef) |
|
|
1494 | PPCODE: |
|
|
1495 | { |
|
|
1496 | SvREFCNT_dec (self->cb_object); |
|
|
1497 | self->cb_object = SvOK (cb) ? newSVsv (cb) : 0; |
|
|
1498 | |
|
|
1499 | XPUSHs (ST (0)); |
|
|
1500 | } |
|
|
1501 | |
|
|
1502 | void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef) |
|
|
1503 | PPCODE: |
|
|
1504 | { |
|
|
1505 | if (!self->cb_sk_object) |
|
|
1506 | self->cb_sk_object = newHV (); |
|
|
1507 | |
|
|
1508 | if (SvOK (cb)) |
|
|
1509 | hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0); |
|
|
1510 | else |
899 | { |
1511 | { |
900 | decode_hexdigit ['a' + i] = 10 + i; |
1512 | hv_delete_ent (self->cb_sk_object, key, G_DISCARD, 0); |
901 | decode_hexdigit ['A' + i] = 10 + i; |
1513 | |
|
|
1514 | if (!HvKEYS (self->cb_sk_object)) |
|
|
1515 | { |
|
|
1516 | SvREFCNT_dec (self->cb_sk_object); |
|
|
1517 | self->cb_sk_object = 0; |
|
|
1518 | } |
902 | } |
1519 | } |
903 | |
1520 | |
904 | json_stash = gv_stashpv ("JSON::XS", 1); |
1521 | XPUSHs (ST (0)); |
905 | } |
1522 | } |
906 | |
1523 | |
907 | PROTOTYPES: DISABLE |
1524 | void encode (JSON *self, SV *scalar) |
|
|
1525 | PPCODE: |
|
|
1526 | XPUSHs (encode_json (scalar, self)); |
908 | |
1527 | |
909 | SV *new (char *dummy) |
1528 | void decode (JSON *self, SV *jsonstr) |
|
|
1529 | PPCODE: |
|
|
1530 | XPUSHs (decode_json (jsonstr, self, 0)); |
|
|
1531 | |
|
|
1532 | void decode_prefix (JSON *self, SV *jsonstr) |
|
|
1533 | PPCODE: |
|
|
1534 | { |
|
|
1535 | UV offset; |
|
|
1536 | EXTEND (SP, 2); |
|
|
1537 | PUSHs (decode_json (jsonstr, self, &offset)); |
|
|
1538 | PUSHs (sv_2mortal (newSVuv (offset))); |
|
|
1539 | } |
|
|
1540 | |
|
|
1541 | void DESTROY (JSON *self) |
910 | CODE: |
1542 | CODE: |
911 | RETVAL = sv_bless (newRV_noinc (newSVuv (F_DEFAULT)), json_stash); |
1543 | SvREFCNT_dec (self->cb_sk_object); |
912 | OUTPUT: |
1544 | SvREFCNT_dec (self->cb_object); |
913 | RETVAL |
|
|
914 | |
|
|
915 | SV *ascii (SV *self, int enable) |
|
|
916 | ALIAS: |
|
|
917 | ascii = F_ASCII |
|
|
918 | utf8 = F_UTF8 |
|
|
919 | indent = F_INDENT |
|
|
920 | canonical = F_CANONICAL |
|
|
921 | space_before = F_SPACE_BEFORE |
|
|
922 | space_after = F_SPACE_AFTER |
|
|
923 | json_rpc = F_JSON_RPC |
|
|
924 | pretty = F_PRETTY |
|
|
925 | allow_nonref = F_ALLOW_NONREF |
|
|
926 | CODE: |
|
|
927 | { |
|
|
928 | UV *uv = SvJSON (self); |
|
|
929 | if (enable) |
|
|
930 | *uv |= ix; |
|
|
931 | else |
|
|
932 | *uv &= ~ix; |
|
|
933 | |
|
|
934 | RETVAL = newSVsv (self); |
|
|
935 | } |
|
|
936 | OUTPUT: |
|
|
937 | RETVAL |
|
|
938 | |
|
|
939 | void encode (SV *self, SV *scalar) |
|
|
940 | PPCODE: |
|
|
941 | XPUSHs (encode_json (scalar, *SvJSON (self))); |
|
|
942 | |
|
|
943 | void decode (SV *self, SV *jsonstr) |
|
|
944 | PPCODE: |
|
|
945 | XPUSHs (decode_json (jsonstr, *SvJSON (self))); |
|
|
946 | |
1545 | |
947 | PROTOTYPES: ENABLE |
1546 | PROTOTYPES: ENABLE |
948 | |
1547 | |
949 | void to_json (SV *scalar) |
1548 | void to_json (SV *scalar) |
950 | PPCODE: |
1549 | PPCODE: |
|
|
1550 | { |
|
|
1551 | JSON json = { F_DEFAULT | F_UTF8 }; |
951 | XPUSHs (encode_json (scalar, F_UTF8)); |
1552 | XPUSHs (encode_json (scalar, &json)); |
|
|
1553 | } |
952 | |
1554 | |
953 | void from_json (SV *jsonstr) |
1555 | void from_json (SV *jsonstr) |
954 | PPCODE: |
1556 | PPCODE: |
|
|
1557 | { |
|
|
1558 | JSON json = { F_DEFAULT | F_UTF8 }; |
955 | XPUSHs (decode_json (jsonstr, F_UTF8)); |
1559 | XPUSHs (decode_json (jsonstr, &json, 0)); |
|
|
1560 | } |
956 | |
1561 | |